Similarity and probability

Abstract Subjects were shown triples of bar graphs, Left, Middle, and Right. One group judged whether L or R was more similar to M. Two other groups were told that the bar graphs described trinomial distributions. Of these groups, one was asked to judge whether sample L or sample R is more likely to emerge from population M. The other group judged whether population L or population R is more likely to yield sample M. All three groups gave essentially the same responses. In particular, the likelihood judgments were closer to the similarity judgments than to the correct likelihoods, in support of the representativeness hypothesis ( Kahneman & Tversky, 1972 ).